8/10/2019 McCloskey Wible Cohen (1988)

1/11

Journal

o

Experimental

Psvchology:

General

1988 Vol. 117,No. 2,

171-'l81

Copyright 1988hythe Am :anPsychological Association, Inc.

OD96-3445/88/$00.75

Is

There a Special Flashbulb-Memory Mechanism?

Michael McCloskey, CynthiaG.

Wible,

andNealJ.Cohen

Johns

Hopkins

University

Many people

reportvivid

recollections

of thecircumstancesin which they

learned

of

major

events, suchas theassassinationofPresident Kennedy,or theassassinationattempton President

Reagan. Brown and Kulik (1977) argued that

this

phenomenon, which they labeledflashbulb

memory,implies

the

existence

of a

special memory mechanism that creates

a

detailed, permanent

record of the individual's experience when triggered by an event exceedingcriterial levelsof

surprise and consequentiality. In this article we evaluate the special-mechanism hypothesis,

arguingon empirical and logical grounds that the flashbulb-memory phenomenon does not

motivate the

postulation

of a special flashbulb-memory mechanism. We suggest instead that

flashbulbmemories should be viewed as

products

of

"ordinary"

memory mechanisms, and hence

as phenomenathatmayoffer insights into the nature of these mechanisms.

Where were youwhenyou learned of the assassination of

John F. Kennedy, the shooting of Ronald Reagan, or the

explosion of the space shuttle

Challenged

How did you hear

about

the event? What were you doing at the time? What

were your first thoughts upon hearing the news? Many people

report

that

they

can

remember

vividlythe

circumstances

in

whichthey learned of certain major public or personal events,

afeatof memory that has intrigued researchers since at least

the turn of the century (e.g., Brewer, 1986; Brown & Kulik,

1977;Colegrove,

1899; Linton, 1975; Neisser, 1982, 1986a,

1986b; Pillemer,1984;Rubin & Kozin, 1984; Thompson &

Cowan, 1986; Winograd&

Killinger,

1983;

Yarmey

&

Bull,

1978).

Brown

and Kulik(1977)suggested that this phenomenon,

whichthey labeledflashbulb memory, implies the existence

ofaspecial memory mechanism. They argued thatthespecial

mechanism, when triggered

by an

event exceeding criterial

levels of surprise and consequentiality, creates a permanent

record of the contents of awareness for the period immediately

surrounding

the

shocking experience.

The

special-mechanism

hypothesis has been the subject of considerable discussion in

recent years, with some authors endorsing

the

hypothesis

and

others noting potential problems. Pillemer

(1984),

for exam-

ple, argued

in

favor

of the

hypothesis

on the

basis

of

data

concerning subjects' recollections of the circumstances in

which they learned about

the

assassination attempt

on

Presi-

dent Reagan. In contrast, Neisser(1982)raised doubts about

the accuracy of flashbulb memories, and also pointed out

problems with several other aspects of the Brown and Kulik

hypothesis. Furthermore, Rubin and Kozin(1984)questioned

whether the

characteristics

of flashbulb memoriesclearlyset

these memories apartfromother autobiographical memories.

Although

a number of important issues

have

been raised,

itremains unclear whether the flashbulb-memory phenome-

We thank Roberta

Goodman-Schulman,

Kathi Hirsh, Brenda

Rapp,Scott Sokol,and Molly Treadway for theirhelpfulcomments.

Wealsothank the Cognitive Lunch group for an illuminating discus-

sion.

Correspondence concerning this article should be addressed to

Michael McCloskey, Cognitive Science Center,

Johns

Hopkins Uni-

versity, Baltimore, Maryland21218.

non warrants the postulation of a special flashbulb-memory

mechanism. One reason is that no detailed, coherent

formu-

lation

of a special-mechanism hypothesis has been set forth.

The hypothesis as stated by Brown and Kulik

(1977)

and

discussed

in

subsequent articles

is

underdeveloped

in

several

crucial respects,

and

there

is

also considerable variation

be-

tween

and

evenwithinarticles

in

what

is

claimed about

the

mechanism and the memories it produces. A related problem

isthat discussions

of the

special-mechanism hypothesis have

oftenfailedto engage the critical issue ofwhatwould consti-

tute

a

sufficientbasis

for

positing

a

special

flashbulb-memory

mechanism.

As

a consequence of these problems, the implications of

the evidence and arguments presented in studies of

flashbulb

memory have not

always

been clear. For example, Neisser

(1982)

reported

two

examples

of

inaccurate flashbulbmem-

ories in support of his argument that these memories are not

necessarily veridical. Thompson and Cowan (1986), however,

presented evidence that for one of Neisser's examples the

inaccuracy is relatively minor, and concluded that this ex-

ample does not challenge the

"basic

accuracy of flashbulb

memories"(p.

200). Neisser

(1986b), in a

reply

to

Thompson

and

Cowan, disagreed with their conclusions.

Unfortunately,

the

exchangefails

to

clarifywhat implications

any

particular

degree

of

inaccuracy

has for the

special-mechanism hypothe-

sis. Do afewexamples of minor inaccuracy call the hypothesis

into serious question? Or does the hypothesis remain unchal-

lengedif the accounts are usually basically accurate ?

In

this

articleweattempttoevaluatethespecial-mechanism

hypothesis systematically and, more generally, to assess the

implicationsof theflashbulb-memory phenomenonforthe-

ories

of

memory.

On the

basis

of

data

and

arguments,

we

conclude

that the postulation of a special flashbulb-memory

mechanism

is unwarranted. Instead, we suggest, flashbulb

memories' should be viewed as products of

"ordinary"

auto-

biographical memorymechanisms,

and

hence

as

phenomena

that mayofferinsights into the nature of these mechanisms.

1

We

willuse the

term

flashbulb memoryin a theoretically neutral

sense, to refer to individuals' recollections of the circumstances in

which they

learned

about surprising and consequential events. It

should

be

understood that

our use of

this term does

not

imply

anythingabout the memory processes underlying the recollections.

8/10/2019 McCloskey Wible Cohen (1988)

2/11

172

M.

McCLOSKEY, C. WIBLE, AND N.

COHEN

The Special-Mechanism Hypothesis

Underlying

the

hypothesis

of a

special

flashbulb-memory

mechanism is the following implicit argument: Flashbulb

memories have

specialcharacteristics,

that

is,

characteristics

different from thosethememories would haveiftheywere

producedby ordinary memory mechanisms. These

special

characteristics imply

that the

memories

are

products

of a

special memory mechanism.

Thus, in evaluating the special-mechanism hypothesis, we

must consider

the

followingquestions: What claims does

the

hypothesis

make

about

the characteristics of flashbulb mem-

ories?

To

what extent

are

these claims warranted

by the

available evidence? And, if these claims are accepted, do they

clearly

distinguish

flashbulb

memoriesfrom"ordinary"mem-

ories and therefore warrant thepostulationof a special flash-

bulb-memory mechanism?

Characteristics of Flashbulb Memories

Statements

of the

special-mechanism hypothesis

are not

entirely clear with regard to the characteristics of flashbulb

memories,

at

timespressing strongclaimswhile

at

other times

qualifying

these claims. As a starting point for discussion, we

will

consideroneparticular interpretationof theBrownand

Kulik (1977) hypothesis, an interpretation that makes strong

claimsabout thecharacteristics of flashbulbmemories.On

this rendering the special flashbulb-memory mechanism,

when triggered by

criterial

levels of surprise and consequen-

tiality,createsinmemoryadetailedandpermanent recordof

the individual's experience immediately before, during, and

immediatelyafterlearning of the shocking event.

(Brown

and

Kulik, 1977, p. 87, stated that the mechanism records all

"brainevents above some leveloforganization, apparently

meaning something like

the

contents

of

awareness, including

such information as where the individual was, what he or she

wasdoing, etc.)

The

representations created

by the

special

mechanism are

complete, accurate,vivid,

and

immune

to

forgetting(i.e.,thestored information,in itsoriginal veridical

form,remains permanently accessible).

One could certainly formulate weaker interpretations of

Brown and

Kulik's

(1977) position (e.g., flashbulb memories

are impressively, although not perfectly, complete, accurate,

vivid,and resistant toforgetting). However,weconsider the

strong claims

first

because these claims,

if

supportable, would

make the most convincing case for the position thatflashbulb

memoriesare soremarkable that aspecial memory mecha-

nism must be posited to account for them. If the strong claims

prove untenable, we can then consider whether weaker claims

motivate

the

postulation

of a

special

flashbulb-memory

mech-

anism.

2

We begin our evaluation of the strong claims about char-

acteristicsof flashbulbmemoriesbyreportingastudy aimed

atevaluating these claims.

Flashbulb Memories for the Explosion

ofthe

Space Shuttle

With one exception (Pillemer, 1984),studiesof flashbulb

memories (Brown & Kulik, 1977;Colegrove, 1899; Winograd

&

Killinger, 1983; Yarmey&Bull, 1978) have probed sub-

jects'memories

for the

circumstances

in

which they learned

of

an eventonlyonce, long

after

the event. Yet,asNeisser

(1982)

has pointed out, this procedure does not permit the

accuracyof the subjects' reports to be assessed. Moreover, it

is

difficult

to determine whether the amount of remembered

information

remains constant over time,

as it

should

if flash-

bulb

memories are immune to forgetting, or whether subjects'

memories instead become less detailed and complete as time

passes.

Potentially more informative

is a

procedure

in

which

sub-

jects are questioned twice, once shortly

after

a surprising,

consequential event, and again some time later, as in

Pille-

mer's

(1984) study concerningtheReagan assassination at-

tempt. The two sets of responses may then be compared with

regard to amount and consistency of

information

reported on

the two occasions. The explosion of the space shuttleChal-

lengeron January 28, 1986, tragically provided the opportu-

nityto carry out a study of this sort. The flight of

Challenger

washighlypublicized because the crew included the teacher

Christa McAuliffe. Furthermore, the explosion was com-

pletelyunexpected and had important consequences for the

space program and, more generally, for public confidence in

government, science,and technology. Thus, with regard to

surprise and consequentiality, the presumed triggering con-

ditions for the special flashbulb-memory mechanism, the

shuttle explosion appears comparable to the events used in

previous studies of flashbulb memory (e.g., the Reagan assas-

sination attempt, assassinations

of

other political leaders).

Indeed, the disaster was widely described in the media as one

of those events for which people remember where they were

and

what they were doing when they learned

of it. In the

present studyweprobed subjects' memoryforcircumstances

of

learning about the shuttle explosion a few days

after

the

explosion,andagain9months later.

Method

Questionnaire. The questionnaire presented to subjects included

four

questionsabout the circumstances under which the

respondent

learned of the explosion of thespaceshuttleChallenger:

1.

Where were you when you first learned of the explosion?

2

It might be suggested that thestrong

claims

we have formulated

about

characteristics

of flashbulb

memories

are

stronger than Brown

andKulik

(1977)

intended. For example, whereas thestrongclaims

characterizeflashbulbmemoriesascomplete recordsof the

individ-

ual's experience around the time of learning about a surprising,

consequential

event, Brown and Kulik (1977, pp.74-75,85)

stated

thatflashbulbmemories varyin"elaboration,"suggestingthatunder

some circumstancesthese memories could be less than complete.

However, Brown

and

Kulik also repeatedlyassert(e.g.,

see

Brown

&

Kulik, 1977,pp. 76,

95-96)

thatthe flashbulb-memorymechanism

recordsallrecent brain events. Thus,potentialconflicts betweenthe

strong claims we have formulated andstatementsmade by Brown

and Kulik

reflect

ambiguities in their claims. In any event, regardless

ofwhat Brown

and

Kulik intended,

it is

important

to

consider strong

claims aboutcharacteristics

of flashbulb

memories because,

as

stated

earlier, these claims wouldprovidethe firmestbasisfor assertingthat

flashbulb

memories

arevery

different from

ordinary memories.

8/10/2019 McCloskey Wible Cohen (1988)

3/11

FLASHBULB MEMORY

7

2. What were you doingwhenyou first learned of the explosion?

3.Did you see theeventat thetimeit wasactuallyhappening,or

did you

learn

aboutit

later?

If

later,

how did you

learn about

it?

4. Whatwereyour first thoughts upon hearing the news?

We will

referto

these

items

collectively

as the

circumstances ques-

tions,andindividuallyas thelocation, activity, source,andreaction

questions, respectively.

For

each question subjects were asked

to

rate

theirconfidence in the correctness of their answers on a scalefrom1

(lowconfidence)

to 7

(high confidence).

The four

circumstances questions were intended

to

probe cate-

gories

of

information that should

be

included

inmemories

generated

byamechanism that recordsthecontentsofawarenessfor theperiod

immediately surroundingthelearning aboutasurprising, consequen-

tial event. Previous studies have focused

on six

canonical

categories"

of information: place, ongoing event, information, own

affect,

affect

in others, and aftermath (see, e.g., Brown &

Kulik,

1977;

Pillemer,

1984).Ourlocation, activity,andsourcequestions closely correspond

to the first

three

of

these categories, respectively,

and our

reaction

question

corresponds roughly

to the ownaffect

category.

We did not

probe affect

in others or aftermath, because information in these

categoriesmay notalwaysbeavailableto berecorded,atleast within

the span

of

time over which

a

special

flashbulb-memory

mechanism

wouldoperate.For

example,

a

person

who

heard

aradio

announce-

ment about the shuttle explosion while alone might have no experi-

ence

ofaffect in

others

to

remember.

The questionnaire alsoincluded

a

single question about

the re-

spondent's memory for his or her whereabouts at the time of the

assassination

attempt

on President

Reagan.

Thesefive

critical questions (the

four

shuttle circumstances ques-

tions

and the

single question about

the

Reagan assassination attempt)

occurred

in

identicalform

on

both

theinitialand

9-month question-

naires. Instructions

on

both questionnaires stressed that answers

should

be as

specific

as

possible.

Both questionnaires also included several additional questions,

most

of

which concerned

the

sources

from

which

the

respondent

obtained information

about

the

shuttledisaster

subsequent

to

initially

learning about it, and the details of theshuttleflightitself.The results

for

these questions are not relevant for our purposes and therefore

will

not be

discussed.

Design

and

procedure.

On January 31, 1986, three days

after

the

explosion, questionnaires were distributed

to 50

faculty,postdoctoral

fellows,graduate

students,

undergraduate students, and support

staff

associatedwith

the

Psychology Department

of

Johns Hopkins Uni-

versity.

Forty-five of these"immediate"questionnaires (90%) were

returnedwithin 1 week of the shuttle disaster.

On October 21,1986,approximately 9 months

after

the explosion,

follow-upquestionnaires were distributed

to the 29

subjects

who (a)

had returned the

immediate

questionnaire

within

1

week

of the

explosion

and (b)

were still available

at the

time

of the

follow-up.

Twenty-seven subjects (93%) completed the 9-month questionnaire.

These subjects

willbe

referred

to as the

repeated

testing

group.

To

assess

effects

of

completing

the

immediate questionnaire

on

responses givenat 9months, the 9-month questionnaire wasalso

distributed to 35 individuals who had not received the immediate

questionnaire. Thirty-one (89%) completed

the

questionnaire. These

subjects

willbe

referred

to as the

9-month-only

group.

Results

Answers

to the

location, activity, source,

and

reactionques-

tions

werefirst

scored

simplyas

substantive

or"don't

remem-

ber responses. Aresponsewasscored assubstantive if it

provided

an

answer

to the

question,even

if the

response

was

nothighly specific.For

example,

at

work

aswellas

Lever-

ingHallnext to the

news

stand wasscoredas asubstantive

answerto thelocationquestion.This scoring criterion appears

to be consistentwith those

used

in previous studies

(e.g.,

Brown &

Kulik,

1977; Pillemer, 1984)toassesswhethera

subjectprovidedinformation

in a

canonicalcategory.Because

some subjects reported

information

relevantto one of the

circumstancesquestionsin an

answer

to another ofthese

questions,

thescoringforeachquestion took accountof any

relevant

information the

subjectprovided

in answers to

other

questions.

Nine-month responses.Table 1presents,for the repeated

testing

subjects

on the immediateand 9-month question-

naires, and for the

9-month-only

subjectson the9-month

questionnaire,theproportionof

subjects

whoofferedsubstan-

tiveresponsestoeachof thecircumstances questions. From

the9-month results,it isclear thattheexplosionofChallenger

is an

event

for which the flashbulb-memory

phenomenon

occurs:9months

after

the

shuttleexplosion

most subjects

could providereports concerningwhere

they

were,

what they

weredoing,howtheyheard,and how

they

reactedwhenthey

learnedof thedisaster.

All

subjects

inboth groups

gave

substantive answersto at

leasttwo of thefour

questions;

22 of the 27

repeated-testing

subjects(81%)and 26 of the 31

9-month-only subjects(84%)

gave substantive

responses

to all

fourquestions. Subjects

were

also

highlyconfident

abouttheirresponses. Acrossthe

four

circumstances

questions,mean confidence forsubstantive

answers was 6.13 (on a

scale

with 7 as the

highest

confidence

level) for the

repeated-testinggroup,

and

6.16

for the

9-month-

only

group.

Brown

and

Kulik (1977)

scored a

subject

as

having

a

flashbulb

memory

for an

event

if the

subject

(a) answered

yes

to the

question,

Do yourecall the

circumstances

in

whichyou first heard

that

[event]? , and (b)

provided

infor-

mation in at least one of the sixcanonical

information

cate-

gories

inreporting these circumstances.Assuming that any

subject

who

gave

asubstantive responseto one ormoreof

ourcircumstances questions

would

have givena yes answer

to aquestionof theform Do you recall the circumstances in

which

you firstheard thatthespace shuttleChallengerhad

exploded? , thenby the

Brown

and

Kulik

criteriaall of our

repeated-testing

and 9-month-only subjects

have

flashbulb

memories for thespace shuttle explosion.

Table

1

Proportion o Repeated

Testing

and

-Month-Only

Subjects

ProvidingSubstantiveResponses to the Circumstances

Questions on theImmediateand9-Month Questionnaires

Repeated testing

Question

Location

Activity

Source

Reaction

Immediate

questionnaire

1.00

1.00

1.00

.96

9-month

questionnaire

.96

.89

1.00

.89

9-month

only

9-month

questionnaire

.94

.87

1.00

.97

.99

.94

.94

8/10/2019 McCloskey Wible Cohen (1988)

4/11

74

M.

McCLOSKEY,

C. WIBLE, AND N. COHEN

Comparison

of the

9-monthresults

for the

repeated-testing

group and the

9-month-only

group suggests that completing

the immediate questionnaire had little if anyeffecton the 9-

month responses of the repeated-testing group. The mean

number of

questions answered substantively

was

3.74 (out

of

4) for the

repeated-testing group

and

3.77

for the

9-month-

only

group, t < \. Furthermore, confidencedid not

differ

across groups (see earlier results on confidencelevels).

Comparison ofimmediateand 9-monthresponses.The 9-

month

results

paint a seemingly impressive picture of subjects'

abilityto remember the circumstances in whichtheylearned

of the

shuttle explosion. However, comparison

of the

9-month

responses with those given shortly

after

the explosion

reveals

that

the

amount

of

stored information subjects could access

decreased over the retention interval, and further that the

immediate and 9-month responses were not always consistent.

For

each

of the 107

circumstances questions answered

substantively on the immediate questionnaire (27 subjects

times 4 circumstances questions, minus a single reaction

question answered "don't remember ),

two

independent

judges compared the immediate response with the corre-

sponding 9-month response.

A

9-month response was scored as the

same

as the corre-

sponding immediate response if it provided basically the same

information(although not necessarily in the same wording).

A

relativelylaxcriterion was applied, so

that,

for example,

theresponses

"reading"

and studying to theactivity ques-

tion

were scored

as the

same.

A9-month response

was

scored

as

morespecific than

the

immediate response if it was not inconsistent with the im-

mediate response, but answered the question at afiner level

of

detail,

or

included particular pieces

of

information

not

reportedon the immediate questionnaire. Forexample,the

9-month response eating lunch (actually, drinking coffee)

with

S S , S D , and M M "to the

activity question was scored as morespecificthan the imme-

diate response eating lunch with some friends.

Similarly, a 9-month response was scored as

moregeneral

than thecorresponding

immediate response

if it was not

inconsistentwiththe immediate response, but was less spe-

cific,

or

omitted particular details.

For

example,

the

9-month

response word of mouth was scored as more general than

the immediate response heardC talking about it.

A9-month responsewasscoredasinconsistentif itcontra-

dicted

the

corresponding immediate response

in

some (al-

though

not necessarily all) respects. For example, the 9-month

response

at my

desk

to the

location question

was

scored

as

inconsistent with the immediate response walking out of the

door toward

B 's

office.

Finally,

a 9-month response was scored asdon'tremember

if

the question was answered substantively on the immediate

questionnairebutgivena"don'tremember responseon the

9-month questionnaire.

The two judges agreed on 86% of the initial classification

decisions. Disagreements were resolved through discussion.

For some ambiguous responses, the subject was consulted for

clarification.

For example, one subject indicated that the 9-

month response examining somedata for an experiment

did not

refer

to thesame activity as theimmediate response

Table 2

ComparisonofImmediateand 9-MonthResponsesfor

Repeated-Testing Subjects

Relationship of 9-month response to

immediate

response

Question

Location

Activity

Source

Reaction

Total

Same

21

15

16

13

65

More

specific

3

2

7

More

general

2

4

7

7

20

Inconsistent

2

4

2

9

Don't

remember

3

0

2

6

designing

an

experiment. Table

2

presents

for

each

of the

fourcircumstances questionsthenumberofresponsesfalling

into each

of the

scoring categories.

Although there

is

considerable consistency over

the 9-

month retention interval, there is also substantial evidence of

forgetting

and inaccuracy.First,the data for"don'tremem-

ber responses clearly reveal forgetting over the retention

interval. On the

immediate questionnaire,

26 of the 27 re-

peated-testing subjects provided a substantive answer to all

fourcircumstances questions,

and the

remaining subject gave

asubstantive response

to

three

of the

fourquestions.

On the

9-month questionnaire, however,4 of the 27 subjects(15.4%)

gave

"don't

remember responses to a total of 6 questions

theyhad answered substantively on the immediate question-

naire.

Further evidence

of

forgetting

may be

found

by

comparing

the

frequency

ofmorespecificandmore general responses.If

the accessible

information

remains constant over time, there

should be no

systematic

difference

between

the

immediate

and9-month questionnaires in

specificity

of responses. That

is,

although

corresponding immediate

and

9-month responses

mayoccasionally varyinspecificitydue tovariationinsub-

jects' decisionsabout

how

much

of the

accessible information

to report, we would expect the number of responses that

became more specificat 9monthsto beapproximatelythe

same as the number that became more general. On the other

hand,if theamountofaccessible

information

decreased over

the interval between questionnaires, we might expect a sys-

tematic

shift

toward more general responses on the 9-month

questionnaire.This

is

just what

we

found. Whereas only

7

responses became more specific,

20

became more general

(p